Designed and developed for production environments
Click to enlarge
Based on a stable storage core, SMTX ZBS 5.0 utilizes advanced hardware technologies and fully-optimized software, providing performance and low latency.
With only a 3-node cluster, it can achieve performance of 770K IO/s for 4K random read. Equipped with enterprise features and intelligent management functionalities, it helps customers build distributed resource pools for large scale virtualization, private cloud, containers, etc.
As a core component of IT infrastructure, SMTX ZBS was born to meet customers’ increasing demand of creating distributed storage resource pools to support existing compute platforms, especially when the expensive and complex legacy storage falls behind the quick pace of digital transformation, and current Ceph-based products are unable to provide high performance and low latency support for mission-critical applications.
Click to enlarge
Based on production-readiness and flexibility of version 4.0, SMTX ZBS 5.0 continues to adapt software and hardware technologies such as Intel Corp.’ Optane Persistent Memory, NVMe-oF, 25GbE/100GbE and RDMA, achieving performance and low latency.
- SMTX ZBS 5.0 introduces RoCE v2 (RDMA over Converged Ethernet) to the storage network to achieve direct communication among the nodes, avoiding OS buffer overhead, which reduces the write latency of remote nodes and improves cluster overall performance.
- Besides standard iSCSI protocol, it supports NVMe-oF, which ensures concurrency and low latency from PCIe devices to the access network, to meet the requirements of performance and low-latency applications.
- It supports an enhanced all-flash solution with Intel PMem as cache and NVMe SSD as storage medium. Leveraging PMem’s performance, low-latency, and non-volatile features, it can lower the latency of business-critical applications.
With above technologies, a 3-node storage cluster of SMTX ZBS 5.0 can achieve 770,000 IO/s for 4K random read; bandwidth can reach 8GB/s for 256,000 sequential read (close to physical limit with a 25GbE network card).